Watch Your Tone: How a Brand's Tone of Voice on Social ...€¦ · HEC Montreal, 3000 Chemin de la Côte-Sainte-Catherine ... Online branding; Digital marketing Introduction The popularity

☆ Funding: This work was supported by the Social Sciences and HumanitiesResearch Council of Canada 430-2015-00621.⁎ Corresponding author at: HEC Montreal, 3000 Chemin de la Côte-Sainte-Catherine, office 4506, Montreal, QC H3T 2A7, Canada.

E-mail addresses: [email protected] (R.H. Barcelos),[email protected] (D.C. Dantas), [email protected] (S. Sénécal).

www.elsevier

https://doi.org/10.1016/j.intmar.2017.10.0011094-9968/© 2017 Direct Marketing Educational Foundation, Inc. dba Marketing EDGE. All rights reserved.

Available online at www.sciencedirect.com

ScienceDirectJournal of Interactive Marketing 41 (2018) 60–80

.com/locate/intmar

Watch Your Tone: How a Brand's Tone of Voice on SocialMedia Influences Consumer Responses☆

Renato Hübner Barcelos ⁎& Danilo C. Dantas & Sylvain Sénécal

HEC Montreal, 3000 Chemin de la Côte-Sainte-Catherine, Montreal, QC H3T 2A7, Canada

Available online 24 November 2017

Abstract

Social media platforms enable firms to communicate directly and often publicly with individual consumers. In this research, comprising fouronline studies, the authors investigate how the tone of voice used by firms (human vs. corporate) influences purchase intentions on social media.Findings suggest that a human tone of voice is not always the firm's best option. Study 1a (N = 174) shows that using a human voice, instead of themore traditional corporate voice, can increase a consumer's hedonic value on social media and also purchase intentions. However, that influence ofa human voice on purchase intentions is stronger when the consumer is looking at a brand page with a hedonic goal in mind (versus a utilitarianone). Study 1b (N = 342) shows that the presence of several negative comments about a brand on social media acts as a boundary condition,nullifying the influence of a human voice on purchase intentions. Studies 2a (N = 154) and 2b (N = 202) show in different settings that using ahuman voice can even reduce purchase intentions in contexts of high situational involvement, due to perceptions of risk associated withhumanness. The results contribute to the literature surrounding the effects of conversational human voice, while also providing managers with a setof guidelines to help inform and identify which tone of voice is best adapted to each communications scenario.© 2017 Direct Marketing Educational Foundation, Inc. dba Marketing EDGE. All rights reserved.

Keywords: Human voice; Social presence; Social media; Online branding; Digital marketing

Introduction

The popularity of social media such as Facebook andTwitter with both consumers and companies has opened upopportunities for new business models and forms of onlinebranding and social commerce. Worldwide spending on socialmedia advertising has increased by 27% from 2015 to 2016,and social media users should reach 2.95 billion by 2020(Statista 2017a, 2017b). However, while social media havebecome an important tool for branding and customer marketing,many questions remain concerning the best ways for brands torepresent themselves or address their customers in this highlyinteractive, both personal and public, conversational environ-ment. Given that social media enable more direct contact with

customers, should a brand present in a more personal and humanway in these contexts? Alternatively, should a brand keep itsdistance and adopt a less intimate approach?

In the professional press, many experts argue that brandsshould use a more human tone of voice on social media (Lundand Sutton 2014). However, there is still little evidence thatthis informal style is the optimal way to communicate withall consumers. Given the reports that brands are increasinglyemploying an informal style in their social media communica-tions (Beukeboom, Kerkhof, and de Vries 2015), the lack ofresearch on its effects on key aspects of consumer–brandrelationships is striking. Even within the same industry, there isno consensus among companies as to the most appropriate tone ofvoice. For example, Visa's brand page on Facebook (@Visa)more often adopts a traditional corporate style of communication,addressing customers using formal language: “Hi [customer].Thank you for your interest in a Visa card. All Visa cards areissued by our client financial institutions. Each one of them has itsown criteria for issuing cards, fees & T&C.” On the other hand,MasterCard's brand page (@MasterCard) adopts a much more

http://crossmark.crossref.org/dialog/?doi=10.1016/j.intmar.2017.10.001&domain=pdf

[email protected]

[email protected]

[email protected]

https://doi.org/10.1016/j.intmar.2017.10.001




61R.H. Barcelos et al. / Journal of Interactive Marketing 41 (2018) 60–80

informal and casual language, expressing emotions and usingemoticons: “Hi [customer]! Do you have a MasterCard? If so, wemay have a small #PricelessSurprise for you☺Hope you have anexcellent long weekend!”

While these differences in communication style are associ-ated with each brand's positioning, they essentially representthe concept of a conversational human voice, which refers toa tone of voice making the company or brand feel closer, morereal and human (Kelleher 2009; Park and Cameron 2014). Thetone of voice concept has deserved some attention in customerservice literature, in which it belongs to the “humanics”category of clues about a product or service (Berry, Carbone,and Haeckel 2002). Such clues tend to address emotionsrather than reason, and are just as important to the customerexperience as the functional clues. When customers deal withfrontline employees or call centers, the tone of voice, andgeneral friendliness of the company's agent becomes a directextension of the brand, and even a personification of thecompany (Brown and Maxwell 2002). However, in interactivemarketing research, the concept of a conversational humanvoice is still a relatively underexplored concept though it hasdeserved growing attention due to evidence in the literaturesuggesting that it increases consumer engagement and brandevaluations (Schamari and Schaefers 2015; Van Noort andWillemsen 2012).

On social media, the brand's tone of voice can be particularlyimportant during initial encounters, when consumers formopinions about new and unfamiliar brands. In such situations,non-verbal cues, such as communication style, play a central rolein reducing uncertainties and influencing assessments of thebrand's trustworthiness (Keeling, McGoldrick, and Beatty 2010).The way brands communicate with consumers can be thusdecisive in shaping consumer attitudes and determining whetherthe relationship will progress beyond the initial encounters(Keeling, McGoldrick, and Beatty 2010). Nevertheless, somestudies have shown that a personal and human communicationstyle can negatively influence consumer attitudes (Gretry et al.2017; Steinmann, Mau, and Schramm-Klein 2015), while othershave found only limited effects (Verhagen et al. 2014). Wesuggest that these contradictory findings may be due to situationalelements that have been overlooked in prior research.

Consequently, the present research aims to fill this gap inthe literature by demonstrating how the adoption of a humanvoice by companies in their social media communicationscan influence consumer response towards the brand. In ourinvestigation, we also consider the effect of interactionsbetween the humanness in the tone of voice used for thebrand and characteristics of the consumption context (namelythe type of consumer goal and level of situational involvement)unaccounted for in previous studies, and which could explainthese disparate findings. As a matter of fact, the type ofconsumer goal and the level of situational involvement areassociated respectively with the hedonic value of the onlineexperience and the perceived risk about the purchase, twoconstructs in consumer behavior known for influencing con-sumer responses towards the brand in online environments(for example, Dawson, Bloch, and Ridgway 1990; Eroglu,

Machleit, and Davis 2003; Pöyry, Parvinen, and Malmivaara2013). Using four online experiments, we show that adopting ahuman (tone of) voice, instead of the more traditional corporate(tone of) voice, can increase a consumer's hedonic value onsocial media and also purchase intentions. However, in thisnotional framework, using a human voice does not increasepurchase intentions for brands in utilitarian contexts (while itdoes in hedonic contexts), and it can even reduce purchaseintentions in contexts of high situational involvement, due toperceptions of risk associated with humanness.

Hence, this research contributes to the incipient and stilllimited body of research in online marketing by exploringconstructs related to humanness in communication on socialmedia using an experimental approach (Gretry et al. 2017;Schamari and Schaefers 2015; Sela, Wheeler, and Sarial-Abi2012). Our results shed some light on the effects of socialpresence and conversational human voice, showing that theeffect of a human voice may be positive, negative or negligibledepending on the consumption context. Furthermore, thisinvestigation has managerial relevance as it informs managerson how to better communicate with consumers on social mediain order to obtain more positive consumer responses. Sincethe products or services offered by the brand may be purchasedto satisfy a hedonic or utilitarian goal, and be associatedwith different levels of situational involvement, the boundaryconditions we tested have practical value. In this sense, weprovide a set of guidelines to help managers identify which toneof voice is more appropriate according to specific contexts.Since social media constitute public environments in whichcommunication between a brand and its customers is visible toall others, any consumer reading the conversations may alsoform an opinion about the brand. Hence, the tone of voice usedby the brand can influence not only the consumers directlyinvolved in the communication on social media but also allother consumers exposed to the conversation.

Human Voice in Online Brand Communications

Even though brands are not human participants per se,consumers can relate and react to them as if they were (Fournier1998). On social media, the brand is always personified to acertain degree because it is an actor interacting with consumerson the same level as any other user, and using a discerniblepattern of communication when talking to them. The stylisticchoices in this pattern of communication – the tone of voice –include attributes such as humanness and closeness that underliethe concept of a conversational human voice (Kelleher 2009).Conversational human voice was originally defined as “anengaging and natural style of organizational communicationas perceived by an organization's public based on interactionbetween individuals in the organization and individuals inpublics” (Kelleher 2009). For the objectives of this research, wefollow a conceptualization of “human voice” similar to Park andCameron's (2014) and define it as a more natural, close, andhuman style of online communication, opposed to “corporatevoice”, which is the more distant and formal style traditionallyused by companies. In practice, companies can use a tone of voice

62 R.H. Barcelos et al. / Journal of Interactive Marketing 41 (2018) 60–80

with any degree of “humanness”, and not only a voice that iseither completely corporate or human.Moreover, even though theconcept of conversational human voice is more often associatedwith the choice of words, the term is not restricted to preciseoperational guidelines (Gretry et al. 2017), referring more broadlyto a “style of organizational communication” (Kelleher 2009).Accordingly, tone of voice “is more than just the words wechoose. It's the way in which we communicate our personality”(Meyer 2016). Thus, what exactly constitutes human or corporatevoice is largely contextual. Beside the choice of words, it can alsoinclude other elements of communication style, such as themusicality in spoken language, or graphic elements in websites,such as emoticons (e.g., Gretry et al. 2017) or the speakers'avatars (e.g., Park and Lee 2013).

In the marketing literature, Schamari and Schaefers (2015)studied conversational human voice as a mediator betweenwebcare and brand–consumer engagement, while Van Noortand Willemsen (2012) studied it as a mediator between webcareand brand evaluations. Other studies have investigated theeffects of communicating in a more human style using differentconcepts associated with human voice such as communicationstyle (Steinmann, Mau, and Schramm-Klein 2015), closeness inlanguage (Sela, Wheeler, and Sarial-Abi 2012), friendliness(Verhagen et al. 2014) or parasocial interaction (Labrecque2014). The idea of a conversational human voice also convergesclosely to social presence context of online communication,as noted by Park and Cameron (2014) and Park and Lee(2013). Social presence theory posits that a critical aspect ofany communication channel is the degree to which the counterpartis perceived as being real in a mediated communication envi-ronment (Short, Williams, and Christie 1976). Since a company'schoice to use a human voice aims to foster the consumer'sperception of the company or brand as being closer, andmore real,from the perspective of social presence theory a conversationusing this tone of voice is also a communication with greatersocial presence.

The importance of communication style and tone of voiceon social media is that it can influence consumer responsestowards the brand, although how and when this happens is notso clear yet. So far, only a few studies have explored conceptsrelated to human voice, or humanized communication, onsocial media in an experimental way; and the findings havebeen dissimilar or inconclusive. For instance, Van Noort andWillemsen (2012) verified that negative brand evaluationsengendered by negative word-of-mouth can be attenuated bymeans of conversational human voice in webcare interventions.However, Verhagen et al. (2014) found only limited effectsof communication style on consumer satisfaction in serviceencounters, and Steinmann, Mau, and Schramm-Klein (2015)found the attitude of members of an online community towardsa brand to be negatively affected when it addressed them in apersonalized communication style. We argue that the underly-ing reason for these inconsistencies might have been the lack ofconsideration for the particular characteristics of each consump-tion situation when testing the effects. Since the consumer'sgoals and needs are very different in each situation, he or shecan respond differently to the use of a human voice by a brand.

For instance, would someone looking for information about ahotel on Facebook and someone posting a complaint about theservices of an internet provider have the same favorable attitudetowards the company if it addressed them using a personal andhuman voice? In other words, there is a possibility of interactionbetween the tone of voice used by the brand and the consumptionsituation that has not been taken into account so far. To addressthis gap of knowledge, we investigate in this paper the effectsof human voice when interacting with two different attributesof the consumption context that moderate its influence on theconsumer: the type of consumer goal (hedonic or utilitarian) andthe level of situational involvement (low or high). We arguethat these attributes of the consumer's context change expecta-tions about a brand and therefore affect her or his response to thebrand when it communicates in a human or corporate tone ofvoice (see Fig. 1).

The Role of Hedonic Value and Consumer Goals

Hedonic value is one of two types obtainable in any con-sumption activity – the other one being utilitarian value –deriving from the emotional benefits and experience of theconsumption itself (Babin, Darden, and Griffin 1994). Researchon online commerce has historically been more interested inutilitarian value, i.e., the rational and instrumental attributes ofwebsites and new media that facilitate online purchases andbrowsing (e.g., Childers et al. 2001; Zhang and Mao 2008).However, more recent studies have shown that the onlineexperience with brands is shaped by the hedonic value theconsumers obtain, for example, from the website's interactivity(Merle, Sénécal, and St-Onge 2012), socialness (Wang et al.2007) and perception of flow (Sénécal, Gharbi, and Nantel2002). On social media in particular, we may expect hedonicvalue to be related to feelings of enjoyment and sociabilityprovided by the connection with the brand or other consumers.Previous works have shown the role of social presence inthe transmission of emotions within a medium (McKenna,Green, and Gleason 2002) even when the presence of the otherparty in the communication is merely perceived (Kumar andBenbasat 2002). As argued before, since a conversation usinga human tone of voice is also a communication with highersocial presence (Park and Cameron 2014; Park and Lee 2013),a brand using a human voice should be able to transmit moreemotions than one using a corporate voice. Hence, we proposethat:

H1. The more human (corporate) the voice used by a brand onsocial media, the higher (less) is the hedonic value of the onlineexperience for the consumer.

Furthermore, we may expect the increased hedonic value tocreate a more favorable disposition of the consumer to marketingstimuli, in a similar fashion to the effect suggested by Van derHeijden (2004) between the perceived enjoyment and the user'sintention to use information systems. In fact, previous studiesin marketing have related the occurrence of pleasant feelingsin retail environments to positive consumer attitude and morefavorable behavior towards the brand (Dawson, Bloch, and

Fig. 1. Research framework.


Ridgway 1990; Pöyry, Parvinen, and Malmivaara 2013). Pöyry,Parvinen, and Malmivaara (2013), for example, have found apositive association between hedonic motivations in online brandcommunities and purchase intentions. Hence, we can expect theuse of a human voice to have a positive influence on consumptionbehaviors towards the brand, such as purchase intentions,because of the increased hedonic value of the online experience(see Fig. 1). Hence:

H2. The influence of the use of a human or corporate voice onpurchase intentions is mediated by the hedonic value of theconsumer's experience with the brand on social media.

However, we can also expect the type of consumption goalto shape the results of the online consumer's experience with aproduct or service, and to moderate the effects of a human voiceon consumer responses. The type of consumer goal (utilitarianor hedonic) emerges from the categorization of utilitarianand hedonic products or services (Hirschman and Holbrook1982). While consumers with utilitarian goals focus on solvingrational needs, and evaluate products according to theirobjective quality, consumers with hedonic goals look mainlyfor enjoyment and evaluate products subjectively (Hassaneinand Head 2006; Hirschman and Holbrook 1982). Even thoughmany products are primarily associated with a specific typeof consumer goal, some can also be purchased to satisfy eitherof them. For example, one can buy a notebook for work(utilitarian goal) or gaming (hedonic goal). Moreover, thetype of consumer goal should not be confused with the typeof value: consumers may get both utilitarian and hedonicvalue in a consumption activity regardless of the type ofgoal (Babin, Darden, and Griffin 1994). In this sense, value is

considered a bidimensional concept (Batra and Ahtola 1990). Aspecific consumer goal, in the other hand, is a unidimensionalcategorization, ranging from purely utilitarian to purely hedonic(Hirschman and Holbrook 1982).

Nevertheless, some studies in electronic commerce haveshown that the attributes most relevant to consumptiondecisions are usually the ones congruent with the consumers'goals. For example, Cai and Xu (2011) show that consumersvalue the expressive aesthetics of an online store (which arehedonic attributes) more when shopping for hedonic productsthan for utilitarian ones. Similarly, Choi, Lee, and Kim (2011)suggest that the level of social presence in online stores hasa greater influence in reuse intentions for hedonic productsthan for utilitarian ones. Given the above, we argue that theinfluence of a human voice on consumer responses on socialmedia should be increased when the consumer is looking at abrand page with a hedonic goal in mind, because the benefitsof a human voice are emotional in nature, and thus morecongruent with a hedonic goal. Conversely, the influence of ahuman voice on consumer responses on social media should bedecreased when the consumer is looking at a brand page with autilitarian goal in mind. Hence:

H3. The influence of the use of a human or corporate voice onsocial media on consumer responses is moderated by the typeof consumer goal, such that that influence is stronger (weaker)when the consumer goal is hedonic (utilitarian).

Since the consumer goal moderates the indirect effect of thetone of voice on purchase intentions, the relationships hypoth-esized by H2 and H3 can be also understood as a moderatedmediation effect (see Fig. 1).


On the other hand, we may also expect the positive influenceof a human voice on consumer responses to be bounded by theoverall valence of the content (posts and comments) on socialmedia, regardless of the type of consumer goal. This is based onprevious work related to the different weighting given byconsumers to negative and positive reviews (for example, Senand Lerman 2007). Even though using a human voice promotesfeelings of sociability and connectedness, the presence ofseveral negative comments about a brand on social mediashould act as a boundary condition, weakening or nullifying theinfluence of a human voice. Besides, as argued by McKenna,Green, and Gleason (2002), the use of a human voice isexpected to increase the transference of emotions betweenparticipants in online communication, and that might includenegative feelings of anger in complaints directed to the brandby unsatisfied customers. Conversely, in the situation of morepositive posts and comments, the use of a human voice shouldincrease the transmission of positive feelings. Hence:

H4. The influence of the use of human voice on consumerresponses is moderated by the overall valence of posts, suchthat the positive influence of human voice on purchaseintentions is weaker (stronger) when posts are mostly negative(positive).

The Role of Perceived Risk and Situational Involvement

The previous hypotheses argue that using a human voice onsocial media may have a positive effect on consumer responsesby means of increased hedonic value. However, we proposethat there are situations in which the consumer should notbe receptive to a more personal and closer tone of voice. Weargue that this should occur in consumption contexts thatincrease the consumer's perception of risk towards the brand.The perceived risk, a well-known concept in the study ofthe consumer decision process, is often conceptualized as theresult of two dimensions: the severity of the possible negativeconsequences of consumer's choice and the uncertainty (orprobability) of occurrence of these consequences (Campbelland Goodstein 2001; Dowling and Staelin 1994; Li, Yang, andWu 2009). Using the example of Campbell and Goodstein(2001), a consumer purchasing a new brand of wine mayperceive risk because she has never tasted it (uncertainty)or because her guests may be displeased if it tastes bad(consequences). Hence, both dimensions contribute jointly tothe perception of risk.

Previous works have shown that, in the presence of a riskyconsumption situation, consumers may engage in certainbehaviors to reduce either the severity of negative conse-quences or their uncertainty. For instance, Campbell andGoodstein (2001) demonstrate that consumers averse to risktend to avoid uncertainty by sticking to familiar options. In thissense, when the consumption situation arouses feelings of risk,they should avoid options that do not correspond to the “norm”.Accordingly, the use of human voice by some brands may beconsidered unusual or atypical if most companies in theirindustry do not communicate in this fashion on social media

or in general. Thus risk-averse consumers might avoid suchsituations. A similar argument is to be found in the servicesmarketing literature. According to Sheth, Newman, and Gross(1991), customers typically evaluate services using two kindsof values: emotional (associated with relational feelings andexperiential benefits) and rational (associated with the satisfac-tion of practical needs). Based on this typology, Li, Yang, andWu (2009), identified that when customers perceive high riskand fear the negative consequences of their choices, theirdecision process focuses mostly on the rational value, i.e., thesolution of the problem at hand. On the other hand, when risk isperceived as low, customers are less concerned with adverseconsequences of their choices and may choose a service basedon the emotional value offered. This rationale applied to theinteraction with brands on social media corroborates the ideathat using a human voice (whose benefits are emotional innature) should have a negative influence in consumer responseto the brand in those consumption contexts where brandscommunicating in a close or informal way are perceived to beriskier.

Situational involvement has been shown to be correlatedwith perceived risk (Huang 2006). Situational involvementdiffers from enduring involvement in the sense that the latteris intrinsic, long-term, and refers to the consumer characteris-tics, while the former is extrinsic, transitory, and refers to theimpact of the service or product consumed on the consumer'swell-being (Huang 2006). Situational involvement is a statein which the consumption problem gains relevance and theconsumer allocates cognitive resources to make better decisions(Houston and Rothschild 1978; Roser 1990). Some workshave noted that the level of situational involvement influencesthe decision process of consumers, especially by exposinguncertainty about the brand or product (Houston and Rothschild1978; Payne, Bettman, and Johnson 1993). Research in onlinecommerce has also shown that situational involvement moder-ates the influence of atmospheric cues of online stores inconsumer evaluations (Eroglu, Machleit, and Davis 2003).

In the context of brand communication on social media, weexpect that the level of situational involvement of the consumershould moderate the influence of tone of voice on consumerresponses. In consumption contexts of high situational involve-ment, the consumer is more sensitive to the outcomes (the properdelivery of a service, for example) and the uncertainty associatedwith the brand. Thus, as argued before, the consumer mayperceive higher risk in brands using a human voice, since this typeof communication is not typical or congruent with the consumer'sexpectations of a company trying to solve a serious matter. Incontexts of low situational involvement, however, the consumeris not so concerned about the outcomes, and the uncertaintyassociated with use of a human voice should not be enough toresult in a high perception of risk. With a low perceived risk, theconsumer should be more open to the emotional benefits of usinga human voice (associated with the hedonic value). Hence:

H5. The consumer's perceived risk towards the brand mediatesthe relationship between tone of voice on social media andpurchase intentions.

1 The exact question was “Check what is the primary aspect you considerwhen choosing a hotel for a (business/vacation) trip. Actually, we are interestedto know whether, or not, participants follow the instructions correctly. To showthat you read the instructions, please check the ‘other’ option in the scale belowand write ‘survey’.” [A list of choices including “price”, “location”, “comfort”,“staff” and “other” followed the question].


H6. The level of situational involvement moderates therelationship between a brand's tone of voice on social mediaand the consumer's perceived risk, such that the use of ahuman voice leads to higher (lower) perceptions of risk and,consequently, lower (higher) purchase intentions, in contexts ofhigh (low) situational involvement. Conversely, the use of acorporate voice leads to lower (higher) perceptions of risk and,consequently, higher (lower) purchase intentions, in contexts ofhigh (low) situational involvement.

Taken together, the relationships hypothesized by H5 andH6 can also be understood as a moderated mediation effect(see Fig. 1).

Study 1a

Stimuli and Procedure

The first experimental study tested the influence of humanvoice on the hedonic value of the online experience (H1), themediation of the effect of human voice on purchase intentions bythe hedonic value of the experience with the brand page (H2),and the moderation of this effect by the type of consumer goal(H3). To test these hypotheses, we chose a 2 × 2 between-subjectdesign (human voice vs. corporate voice × hedonic consumergoal vs. utilitarian consumer goal). We chose Facebook as thesocial medium for the studies because of its widespread adoptionand relevance to business today (Snyder 2015). Facebook brandpages are also a type of social media where consumers can seeconversations with the brand and among themselves, a featurethat is useful for our purposes.

We created versions of a fictitious hotel brand page forthe study. The use of a fictitious brand had the objective oftesting participants' reactions to human voice without biasesfrom previous attitudes towards brands. We chose a hotelas the context of study given the importance of relationalaspects in service evaluation such as care and responsiveness(Bitner, Booms, and Tetreault 1990). The participants wererecruited using Amazon MTurk and received a link to thesurvey (206 participants, U.S. residents excluding those fromNYC). Following the recommendations of Peer, Vosgerau, andAcquisti (2014) for data quality, we restricted participation toMTurk workers with high reputation (above 95% approvalratings, minimum 500 HITs). They were randomly assignedto one experimental condition and instructed to imaginethemselves going on a trip to New York City and looking onFacebook for opinions about a hotel named “The Whitaker HotelNew York”. The purpose of the trip depended on the assignedconsumer goal condition: either for business (utilitarian) orvacation (hedonic). Then, each participant was shown one of twoversions of the hotel brand page on Facebook (using a humanor corporate voice). After looking at the pages for as long as theywished, participants responded to a survey incorporating themeasures of interest in the study.

To improve the quality of the answers in our sample, weimplemented some control measures. First, we inserted JavaScriptcode in the survey to allow access only to laptops and desktops,

thus avoiding the problem of having participants look at reducedversions of brand pages on the small screens of mobile devices.Second, we included a hypothesis-guessing question at the endof the survey to assess what participants thought was the purposeof the research. Responses indicated that participants were notcognizant of the hypotheses. Third, in order to avoid using rushedanswers that did not actually reflect participants' opinions, weincluded an attention question1 on the survey and removed fromthe sample the answers from two participants who did notcomply. Finally, we also inserted a hidden timer on the onlinesurvey that measured how long the participants looked at thebrand pages. Then, we discarded from the sample the answersof those who had not looked at the pages for at least 15 s(27 participants). 15 s was the minimum time measured in thepre-tests for the participants to make an initial evaluation aboutthe brand.

Tone of Voice ManipulationThe two versions of the hotel brand page were exactly the

same, except for the specific manipulations on human orcorporate voice (see Table 1 and extracts in Appendix A).These manipulations were inspired by previous work (Park andCameron 2014; Park and Lee 2013; Steinmann, Mau, andSchramm-Klein 2015) and consisted of choices of languageand pictures that should increase (vs. decrease) the perceptionof the brand as being more human and close. To improverealism, all posts and replies were based on the content from thebrand pages of real NYC hotels, however with names, pictures,and personal details changed. Both pages had 29 posts, 11being initiated by the brand and 18 by users; 10 posts containedpictures and 9 were text-only; 13 posts were primarily positive(compliments by customers or self-promotion by the hotel),6 were primarily negative (complaints by customers) and 10were neutral (questions and answers).

Consumer Goal ManipulationWe manipulated the consumer goal through the instructions

given to participants before looking at the brand pages. In theutilitarian goal scenario, participants were instructed to imaginethemselves in need of a hotel for a business trip. In the hedonicgoal scenario, they were instructed to imagine themselvesgoing on a vacation trip. Use of the same basic service (hotel)for the two types of consumer goal had the objective ofavoiding possible confounding effects of idiosyncratic differ-ences between two unrelated services (Kronrod and Danziger2013). In this sense, the choice of New York as the city forboth scenarios was also useful because the city is a traditionaldestination for both business and vacation travels. As primingquestions for the consumer goal manipulation, we also askedparticipants which attributes of hotels were more relevant tothem on a business or vacation trip.

Table 1Tone of voice manipulation in the studies a.

Tone of voice Manipulations Examples

Human Brand page uses the employee's avatar as the profile picture. Posts useinformal language and express more often emotions like happiness orsadness. The author addresses consumers by their first names, refers toitself mostly in the first person (“I” or “we”) and signs with his/herown name.

“Thank you very much, Marie! We're so happy to hear that! ☺”“Hello, Marta. We are very sorry for the delay in communication.”“I will attempt to expedite your request. Thank you, Kate.”

Corporate Brand page uses the brand logo as the profile picture. Posts use formallanguage and avoid the expression of emotions. The author addressesconsumers only by their last names or pronouns, refers to itself mostlyin the third person (“the hotel”, “the staff”) and signs with thebrand name.

“The Whitaker Hotel appreciates your kindly words, Ms. Harrington.”“Ms. Harmon, we apologize for the delay in communication.”“The staff will attempt to expedite your request. Thank you, TheWhitaker Hotel.”

a The manipulations were inspired by previous works relating language choices and feelings of closeness/humanness (Fournier 1998; Park and Cameron 2014; Parkand Lee 2013; Sela, Wheeler, and Sarial-Abi 2012; Steinmann, Mau, and Schramm-Klein 2015).


MeasuresWemeasured purchase intentions by assessing the participants'

likelihood of staying at the hotel for a [business/vacation] tripin a 7-point Likert type scale. To measure hedonic value and tocheck for the manipulation of human or corporate voice, we usedscales adapted from the literature (see tables in Appendix A).For control variables, besides the usual demographic variables(gender, age, and education), we included the measure of theparticipants enduring involvement with hotels, the participants'expertise in hotels, and the participants' Facebook usage intensity(see scales in Appendix A).

Pre-testsWe refined and improved the stimuli and study measures by

means of three pre-tests. In the first pre-test (101 participants),we tested brand pages with different numbers of posts (15 or30 posts) to determine the appropriate amount that the participantsshould read in order to exhibit different reactions to the tone ofvoice used. We considered the longer version more appropriateto our manipulations. In the second pre-test (52 participants),we tested all measures as well as the average perceptions ofrealism (M = 5.85 out of 7) and role-playing (M = 6.10 outof 7) and considered them satisfactory. In the final pre-test(63 participants), we added new refinements such as the codeprohibiting the visualization of the brand pages on mobiledevices, and removed one post considered too negative andpotentially biasing purchase intentions for some participants.We then used this final version of the scenarios with 29 posts inthe main study.

Results

The final sample (n = 174) comprised 88 men aged 34.5 onaverage. The manipulations of tone of voice (α = .91) andconsumer goal were successful (Morg = 4.98; Mhuman = 6.30;F(1, 170) = 120.10, p b .001; Mbusiness = 3.46; Mvacation = 4.76;F(1, 170) = 44.28, p b .001). As in the pre-tests, perceptionsof realism (M = 5.79) and role-playing (M = 5.80) weresatisfactory. We did not observe any significant differencesin the means of the dependent variables in terms of gender,age, or education. However, enduring involvement with hotels(α = .95; M = 2.88) was a significant covariate for purchase

intentions (β = −.17; F(1, 168) = 7.20, p b .01) and hedonicvalue (β = −.32; F(1, 168) = 27.82, p b .001). Facebook usageintensity (α = .89; M = 3.54) was also a significant covariate forpurchase intentions (β = .25; F(1, 168) = 9.14, p b .01) andhedonic value (β = .28; F(1, 168) = 10.31, p b .001). The signof the parameter estimates suggests that consumers with lowerenduring involvement with hotels and heavy users of Facebookare more likely to make a reservation at the hotel and experiencemore hedonic value on the brand page.

We observed a significant main effect of tone of voiceon hedonic value (α = .89; Morg = 4.25; Mhuman = 5.44;F(1, 168) = 36.87, p b .001), supporting H1. Results fromMANCOVA analysis also showed a significant main effectof tone of voice on purchase intentions (Morg = 4.87;Mhuman = 5.83; F(1, 168) = 25.17, p b .001) and a significantmain effect of the type of consumer goal (Mbusiness = 5.21;Mvacation = 5.59; F(1, 168) = 4.78, p b .05). More importantly,the results also show a significant effect of the interactionbetween tone of voice and consumer goal on purchase intentions(F(1, 168) = 4.47, p b .05; η2 = 0.03; see Fig. 2). The useof human voice increased hotel reservation intentions more inthe case of a hedonic consumer goal (vacation travel) thana utilitarian consumer goal (business travel). Hence, H3 wassupported.

To test whether the hedonic value of the consumer's onlineexperience mediated the relationship between tone of voice andpurchase intentions, we ran a mediational analysis using thePROCESS SPSS macro (Hayes 2013; model 14 with 5,000bootstrapping samples). In the regression model, the depen-dent variable was purchase intentions, while the independentvariable was tone of voice, the moderator was type of consumergoal, and the mediator was hedonic value (Fig. 3). Involvementwith hotels and Facebook usage intensity were covariates. Theeffect of the mediator on the dependent variable was significant(β = .29, SE = .08, t = 3.49, p b .001) and the interactiveeffect of hedonic value and type of consumer goal was alsosignificant (β = .27, SE = .13, t = 2.15, p b .05). ConfirmingH2, the index of moderated mediation (index = .12, SE = .06,95% CI [.02, .25]) indicated that the mediating role ofhedonic value in driving the effect of tone of voice on purchaseintentions is moderated by the type of consumer goal. When theconsumer was looking for a hotel for a vacation trip, the hedonic

1

2

3

4

5

6

7

Corporate voice Human voice

Hot

el r

eser

vati

on in

tent

ions

Business travel (utilitarian goal) Vacation travel (hedonic goal)

Fig. 2. Interactions between brand voice and consumer goal in Study 1a.


value of the online experience mediated the effect of tone ofvoice on purchase intentions (indirect effect = .19, SE = .05,95% CI [.10, .31]). This was not the case, however, when theywere looking for a hotel for a business trip (indirect effect = .07,SE = .05, 95% CI [−.02, .17]).

Discussion

The results of Study 1a support our first three hypotheses.The brand's use of human voice on social media increased thehedonic value of the online experience for the consumer (H1)and, as a result, also increased purchase intentions (H2).As hypothesized, the type of consumer goal moderated thiseffect (H3). The results also corroborate our reasoning for thisinteraction, i.e., the congruence between hedonic value andthe type of goal. In addition, the findings show that heavyFacebook users and consumers with lower enduring involve-ment with hotels were more likely to make a reservation atthe hotel and experience more hedonic value on the brandpage. The positive influence of Facebook usage intensity wasexpected, since this variable represents the importance of thissocial medium in people's lives. Thus, intense Facebook usersshould get more hedonic value when reading brand pages

Fig. 3. Moderated mediatio

and are likely to give more importance to the posts they read.The negative influence of the level of enduring involvement,however, apparently contradicts the findings of Wang et al.(2007) on the relationship between hedonic value, involvement,and perceived social character of websites. Perhaps consumersmore involved with hotels have greater knowledge for making adecision about a reservation on their own, and are less easilypersuaded by peripheral action of the brand, such as the use ofhuman voice. Study 1b expands this investigation and tests ifthe presence of overall negative posts also limits the influenceof human voice, as hypothesized in H4.

Study 1b


In the second study, we tested the moderating influenceof the overall valence of posts in the consumer responses (H4).A 2 × 2 × 2 between-subject design (human voice vs. corpo-rate voice × hedonic consumer goal vs. utilitarian consumergoal × positive vs. negative overall valence of posts) was usedto test the hypotheses. The experimental procedure was similarto the previous study. We recruited a sample of participantsfrom Amazon MTurk (401 participants respecting the samecriteria used for Study 1a) and instructed them to imaginethemselves looking for a hotel in New York City. Again, theywere randomly assigned to the condition of human or corporatevoice on the brand page and to the condition of business orvacation travel (utilitarian or hedonic goal). Then they hadto report their likelihood of making a reservation at the hoteland answer the other measures of interest in the study.However, in order to test the influence of the overall valenceof posts, we also designed brand pages with more positive ormore negative posts. Hence, we used four versions of the hotelpage in this study (human voice vs. corporate voice × positivevs. negative overall valence of posts). In comparison, whilethe brand pages with positive overall valence had 13 positiveposts, 10 neutral posts and 6 negative posts, the brand pageswith negative overall valence had 3 positive posts, 9 neutralposts and 17 negative posts. Positive posts contained primarilycompliments from consumers, neutral posts contained general

n model for Study 1a.

A. Overall positive posts

2

3

4

5

6

7

Hot

el r

eser

vati

on in

tent

ions


information and questions, and negative posts contained mainlycomplaints from clients. Other than that, the number of likes,reviews and info about the hotel remained the same. We alsokept an approximately equal number of replies by post, postsinitiated by the brand or by customers, and posts with picturesor text-only. As before, all posts and replies were adapted fromextracts of real brand pages of hotels in New York City. Themanipulation of human voice followed the same principle as inthe previous study. Manipulation of the consumer goal was alsothe same, as were the general measures and controls taken toimprove internal validity such as the attention questions and thehidden timers on the brand pages. The general measures in thestudy were also the same.

B. Overall negative posts

1Corporate voice Human voice


3

4

5

6

7ot

el r

eser

vati

on in

tent

ions

Pre-testThe pages with positive overall valence of posts were the

same used in Study 1a. We then designed the new pages withnegative overall valence of posts with the criterion that overallperceptions should be considerably lower than in the pagesused in the previous study, but not so low as to saturate themeasure of purchase intentions at the bottom of its scale. Apre-test (n = 42) for the new brand pages showed that averagepurchase intentions across experimental conditions (M = 3.24out of 7) were indeed lower than those of Study 1a, but stillin the middle-range of the scale. Moreover, means of realism(M = 5.57) and role-playing (M = 5.51) were satisfactory.Hence, we considered the new brand pages with negativeoverall valence of posts appropriate for the main study.

1

2


H


Fig. 4. Interactions between brand voice and consumer goal in Study 1b.

Results

As in Study 1a, we removed participants who remained onthe brand page for less than 15 s (n = 49) or who failedon the attention question (n = 10). Thus, the final sampleconsisted of 342 participants, of which 187 were men, with anaverage age of 35.9. Again, the manipulations of tone of voice(α = .91) and consumer goal were successful (Morg = 4.89;Mhuman = 5.56; F(1, 340) = 25.67, p b .001; Mbusiness = 3.58;Mvacation = 4.50; F(1, 340) = 39.62, p b .001). Perceptionsof realism (M = 5.79) and role-playing (M = 5.96) of thescenarios were satisfactory. We did not observe any significantdifferences in the means of the dependent variables in terms ofthe covariates.

Results from ANOVA analysis showed a significant effectof tone of voice on hotel reservation intentions (Morg = 4.02;Mhuman = 4.34; F(1, 334) = 4.873, p b .05), but also an effectfrom the interaction between tone of voice and overall valenceof posts (F(1, 334) = 4.92, p b .05; η2 = 0.02). Analyzing theconditions of overall positive and negative posts separately, wefound a positive effect of human voice on hotel reservationintentions when posts were mostly positive (Morg = 4.72;Mhuman = 5.45; F(1, 164) = 12.246, p b .001), but no signifi-cant effect when posts were mostly negative (Morg = 3.32;Mhuman = 3.32; F(1, 170) = .000, p = .99; see Fig. 4). Thissuggests that the positive effect of human voice on thedependent variable (also seen in Study 1a) vanished when posts

on the brand page were mostly negative. Hence, the findingssupport H4.

Discussion

Because of the negative opinions about the brand, purchaseintentions were lower for the participants who looked at thebrand pages with overall negative posts, which was evidentlyexpected. More importantly, however, the difference betweenpages using human voice or corporate voice became insignif-icant in the experimental conditions of negative overall valenceof posts. In other words, the overall negative posts about thebrand on those pages decreased the influence of the tone ofvoice on the brand page and actually made it insignificant.Thus, when posts on the brand page are mostly negative, brandscould opt for either a human or a corporate voice in theircommunication, since the overall effect is not significant. Whenposts are mostly positive, brands can opt for a human voice, asit increases purchase intentions.


Study 2a


The objectives of this study were to test the mediation of theeffect of tone of voice on purchase intentions by the perceivedrisk about the brand (H5) and the moderation of this effect bythe level of consumer's situational involvement (H6). We useda 2 × 2 between-subject design (human voice vs. corporatevoice × low SI vs. high SI) to test these hypotheses. Wechanged the context from hotels to restaurants to test theinfluence of the human voice in a different setting and thereforeincrease external validity. The use of a restaurant also servedfor manipulating the situational involvement, since a dinnermay be more or less important for customers depending onthe extrinsic meanings associated with the occasion. Thus, wecreated two versions of the Facebook brand page (using humanand corporate voice respectively) for a fictitious Italianrestaurant named Francesco's Pizza & Pasta. As in the previousstudies, we recruited the participants using Amazon MTurk(200 participants with more than 500 HITs, 95% minimumapproval, not participating in previous studies, U.S. residentsonly) and directed them to our online survey. Participants wereinstructed to imagine themselves in the situation where theywere looking for a restaurant for dinner and found the brandpage of Francesco's restaurant on Facebook. They were askedto examine the brand page freely until they had formed anopinion about the restaurant, and then proceed to the survey.The main dependent variable in this survey was the likelihoodof them choosing to have dinner at this restaurant on thatspecific occasion.

Human Voice ManipulationThe manipulation of human voice followed the same

principles as in the previous studies. Accordingly, the profilepicture on the brand page using a corporate voice displayed thelogo of Francesco's Pizza & Pasta, while the profile pictureon the brand page using human voice displayed a picture ofthe restaurant's chef, and who could be regarded as Francescoitself. Both versions of the brand page had 16 posts and thesame number of likes and reviews. The contents of the brandpages were the same, except for the manipulations in voicetype. To improve realism, all posts and replies were based onreal brand pages from real restaurants, but with names, pictures,and personal details changed.

Manipulation of the Situational InvolvementWe manipulated the situational involvement by means of the

instructions given to participants before looking at the brandpages. In the low situational involvement scenario, participantswere told to imagine themselves searching for a restaurantto have a simple, casual dinner alone. In the high situationalinvolvement scenario, participants were told to imagine them-selves in need of a restaurant to have a birthday celebration withtheir family and friends. As before, the use of the same basicservice (restaurant) for the two situational involvement scenarioshad the objective of avoiding the possible confounding effects

of idiosyncratic differences between two unrelated services(Kronrod and Danziger 2013). In this sense, we considered thatan Italian restaurant would be suitable for dinners of both low-and high-levels of situational involvement.

MeasuresWe reused some measures from the previous studies such

as the manipulation check for tone of voice and the controlvariables. We measured purchase intentions by assessing theparticipants' likelihood of going to the restaurant for a “casualdinner alone” or a “birthday dinner with your family and friends”with a 7-point Likert type scale. In addition, we measuredperceived risk and checked for the manipulation of the situationalinvolvement using scales adapted from the literature (see tables inAppendix B).

Pre-testsThe stimuli and measures of the study were refined and

improved after two pre-tests. In the first pre-test (n = 49), weproposed six different occasions for dinner in a restaurantand selected the two which had scored the lowest and highestin the situational involvement scale. In the second pre-test(n = 52), we tested the final versions of the brand pages andverified satisfactory means for realism (M = 5.62) and role-playing (M = 5.84). We also tested if the manipulationsof situational involvement would also have an unpredictedeffect on participants' understanding of consumer goal andfound no significant differences (Mlow = 4.52; Mhigh = 5.18;F(1, 51) = 2.29; p = .14). Hence, we tested an independenteffect in this study.

Results

The final sample consisted of 154 participants (39 participantswho remained less than 15 s on the brand page and 7 participantswho failed on the attention question were eliminated), of which77 were men with an average age of 34.3. The manipulations oftone of voice (α = .91) and situational involvement (α = .86)were successful (Morg = 4.86; Mhuman = 5.64; F(1, 150) =20.86, p b .001; Mlow = 5.05; Mhigh = 5.69; F(1, 150) = 13.15;p b .001). The average perceptions of realism (M = 5.73) androle-playing of the scenarios (M = 5.99) were satisfactory. Wedid not observe significant differences in the means of thedependent variables in terms of any of the covariates in thestudy. In order to further check for the independence of themanipulations of situational involvement in this study fromthe manipulations in consumer goal in the previous ones, we alsomeasured the hedonic value of the online experience. We foundno significant differences in terms of situational involvement(Mlow = 4.77; Mhigh = 4.86; F(1, 152) = .28; p = .60).

The results fromMANOVA analysis did not show significantmain effects of tone of voice (F(1, 150) = .00, p = .99) orsituational involvement (F(1, 150) = .05, p = .83) on purchaseintentions. However, we observed a significant effect of theinteraction between tone of voice and the level of situationalinvolvement on the dependent variable (F(1, 150) = 6.43,p b .05; η2 = .04; see Fig. 5). In the low situational

1

2

3

4

5

6

7


Pur

chas

e in

tent

ions

Casual dinner (low SI) Celebration dinner (high SI)

Fig. 5. Interactions between brand voice and situational involvement in Study 2a.


involvement condition, the use of human voice increasedpurchase intentions, while in the high situational involvementcondition, it decreased. Hence, H6 was supported. In the sameway, the results did not show significant main effects ofthe tone of voice (F(1, 150) = .00, p = .95) or situationalinvolvement (F(1, 150) = .33, p = .56) on consumers' per-ceived risk (α = .92), but they showed a significant effectfrom the interaction between the two independent variables(F(1, 150) = 6.46, p b 0.05; η2 = .04).

To test the moderated mediation hypothesis, we rana mediational analysis using the PROCESS SPSS macro(Hayes 2013; model 7 with 5,000 bootstrapping samples). Inthe regression model, the dependent variable was purchaseintentions, while the independent variable was tone of voice,the moderator was level of situational involvement, and themediator was perceived risk (Fig. 6). The effect of the mediatoron the dependent variable was significant (β = −.60, SE = .07,t = −8.57, p b .001). The interactive effect of tone of voice andlevel of situational involvement on perceived risk was alsosignificant (β = 1.19, SE = .48, t = 2.49, p b .05). ConfirmingH6, the index of moderated mediation (index = −.71, SE = .31,95% CI [−1.38, −.16]) indicated that the mediating role ofperceived risk in driving the effect of tone of voice on purchase

Fig. 6. Moderated mediatio

intentions is moderated by the level of situational involvement.When the consumer was looking for a restaurant for a birthdaydinner with friends, the perceived risk mediated the effect oftone of voice on purchase intentions (indirect effect = −.36,SE = .19, 95% CI [−.79, −.02]). The event evoked a highperception of risk, and this increased risk decreased purchaseintentions. This was not the case, however, when they werelooking for a restaurant for a casual dinner (indirect effect = .35,SE = .22, 95% CI [−.06, .81]).

Discussion

The results of this study confirmed our hypotheses. In thelow situational involvement scenarios, the use of human voicehas a positive influence on purchase intentions (as in Study 1a).In the high situational involvement condition, however, theincreased sensitivity of consumers to risk was enough toreverse the influence of human voice and make it negative.Consumers in this situation preferred the more distant andformal voice on the brand page. The mediation analysis showedthat the perception of risk mediated the relationship betweenpurchase intentions and the interaction of human voice andsituational involvement. The next study tests that interactionagain but with a different approach.

Study 2b


This study used a different experimental setting to increase thevalidity of the findings about the moderation of the relationshipbetween tone of voice and purchase intentions on social media bythe consumer's situational involvement (H6). In this study, wepresented two different brands and asked the participants tochoose one of them. Moreover, we used a product (wine) insteadof a service to increase external validity. We used a 2 × 2between-subject design (use of human voice by brand A or bybrand B × low SI vs. high SI) to test the hypothesis.

The setting for the online experiment was the choice of abrand of wine brand for a specific consumption situation. Wineis an experiential product for which consumers usually look for

n model for Study 2a.


other people's opinions and brand attributes in their decisionprocess (Sénécal and Nantel 2004), so we considered it anadequate choice for our experiment. The two fictitious brandpages of wineries created for the study (Valmont and Grenier)were based on actual posts from Californian wineries whosewines were included in a list of best affordable Americanwines. In order to avoid confounded effects of perceptions ofquality, origin, and price, the two wine labels reflected designsof wines within the same price range and included the samevarieties of red and white wine. We told participants to imaginethemselves in the scenario in which they needed to choose abrand of wine for a given occasion and were then presented thebrand pages for two Californian wines. They then had to chooseone of the two wines and answer other measures of interest inthe survey. Since the participants could examine each pagefreely, we included independent timers in both pages. Thepresentation order of the pages (Valmont or Grenier first) wasrandom but recorded and included in the analysis to check forpossible effects. As in the previous studies, the participants wererecruited using Amazon MTurk (251 participants respecting thesame criteria used for the previous study) and directed to oursurvey with the experimental stimuli.

Human Voice ManipulationThe manipulation of human voice followed the same

principles as in the previous studies, adapted however to thebrand pages of Valmont and Grenier wineries. Participantswere presented with the version using a human voice on one ofthe pages and the version using a corporate voice on the other.Depending on the experimental condition randomly assigned toeach participant, the brand page using human voice would beeither Valmont or Grenier. The two versions of the brand pageshad 11 posts (6 positive, 4 neutral, and 1 negative) and theircontents (post length, number of replies, and comments) weresimilar, except for the tone of voice manipulation (see extractsin Appendix C).

Situational Involvement ManipulationAs in the previous study, we manipulated the situational

involvement by means of the instructions given to participantsbefore looking at the brand pages. In the low situationalinvolvement scenario, we told participants to imagine them-selves looking for a brand of wine to taste alone at home. In thehigh situational involvement scenario, they were told toimagine themselves looking for a brand of wine as a birthdaypresent for someone special who loves wine.

MeasuresOur dependent variable in this study was the brand chosen

by each consumer (Valmont or Grenier). We reused somemeasures from the previous study such as the manipulationcheck for tone of voice and the control variables. We alsoincluded as control variables the order of presentation ofthe brand pages and the measure of the consumer's winepurchasing habits (by asking the estimated number of bottlesbought per month).

Pre-testThe stimuli and measures of the study were refined and

improved after a pre-test (n = 78) in which participants wouldrandomly look at only one of the brand pages. The objectivehere was to evaluate individually the tone of voice manipula-tion (α = .92) in each version of the brand pages. In this sense,the difference in the means of tone of voice between the pagesusing corporate (MValmont = 5.07; MGrenier = 5.02; Maverage =5.05) and human voice (MValmont = 5.60; MGrenier = 5.62;Maverage = 5.61) was significant (F(1, 74) = 4.45; p b .05),while the difference in the means between Valmont and Grenierwas not, which was expected.

Results

The final sample consisted of 202 participants (42 partici-pants who remained less than 15 s on any of the brand pagesand 7 participants who failed on the attention question wereeliminated), of which 98 men with an average age of 37.7.Again, the manipulation of situational involvement (α = .95)was successful (Mlow = 4.43; Mhigh = 5.78; F(1, 198) = 40.89;p b .001). Since the manipulation check of human voice wasalready successful in the pre-test, we did not include it inthe final study because it would likely suffer a bias fromparticipants' brand choice, i.e., they could evaluate each brandin the sense of forcedly confirming their previous choice.Perceptions of realism (M = 5.81) and role-playing (M = 5.89)were satisfactory. The order of presentation of the brand pages(random in the study) was a significant covariate for brandchoice (β = .29; 95% CI [.00, .57], p = .05), with the valence ofthe parameter estimates suggesting that consumers were a littlemore likely to choose the first brand of wine presented to them(i.e., primacy effect, Carlson, Meloy, and Russo 2006). We didnot observe any significant differences in the means of thedependent variables in terms of the other covariates.

To investigate if brand choice is influenced by the interactionbetween human voice and the level of consumer's situationalinvolvement, we ran a logistic regression using the PROCESSSPSS macro (Hayes 2013; model 1 with 5,000 bootstrappingsamples). The voice used by Valmont (human coded as 1,corporate coded as −1) and the level of situational involvement(high coded as 1, low coded as −1) were the independentdiscrete variables, while brand choice was the dependentdiscrete variable (Valmont coded as 1 and Grenier as 0). Thevoice used by Grenier was the reverse of Valmont's by design,so it was not included in the model. The order of presentation ofthe brand pages was a covariate. Confirming our expectations,the interaction effect between tone of voice and situationalinvolvement was significant in the model (β = −.30, SE = .15,z = −2.07, p b .05), while the main effects of tone of voice(β = −.04, SE = .15, z = −.27, p = .79) and situational in-volvement (β = −.27, SE = .15, z = −1.82, p = .07) were not.In the condition of low situational involvement, the number ofparticipants who chose Valmont increased when this brand wasthe one using human voice (44.90% Valmont using corporatevoice and 55.10% Grenier using human voice; vs. 56.60%Valmont using human voice and 43.40% Grenier using


corporate voice; see Fig. 7). However, in the condition of highsituational involvement, the number of participants who choseValmont decreased when this brand was the one using humanvoice (46.94% Valmont using corporate voice and 53.06%Grenier using human voice; vs. 31.37% Valmont using humanvoice and 68.63% Grenier using corporate voice; see Fig. 7).Hence, the results of the study also supported H6.

Discussion

The findings of this study again verified that the level ofsituational involvement moderates the effect of the use ofhuman voice on purchase intentions. The consumers' prefer-ence between the two brands shifted, in the sense that the brandusing human voice was favored when situational involvementwas low, but the brand using corporate voice was favored whensituational involvement was high. Evidently, the influence of ahuman voice in consumer decisions was magnified in this study

A. Low situational involvement

B. High situational involvement

0%

10%

20%

30%

40%

50%

60%

70%

80%

Valmont usescorporate voice

Valmont useshuman voice

Bra

nd c

hoic

e

Valmont Grenier

0%

10%

20%

30%

40%

50%

60%

70%

80%

Valmont usescorporate voice

Valmont useshuman voice

Bra

nd c

hoic

e

Valmont Grenier

Fig. 7. Interactions between brand voice and situational involvement inStudy 2b.

in comparison to real-life situations, since all other factors thatcould influence consumer decision in that specific case werecontrolled. Nevertheless, it is important to note that a humanvoice can still influence product evaluations, as shown in thisstudy, even though we might expect that influence to be greaterin the evaluation of services due to the importance of relationalaspects (Bitner, Booms, and Tetreault 1990).

General Discussion

In this paper, we analyzed how the use of human orcorporate voice in brand communication influences consumers'responses towards the brand on social media. More specifically,we analyzed this influence in different contexts in which thetone of voice may have a positive or negative effect on aconsumer's responses depending on its interaction with theconsumer's context (namely, his/her consumer goal and his/hersituational involvement with the service). From a theoreticalcontribution perspective, our investigation responds to a gap inthe electronic commerce literature, i.e., the lack of constructsconcerning the social aspects of communication betweenconsumers and companies (Liang and Turban 2011). Theconcept of human voice is still underexplored in the marketingliterature and it can address the effects of strategies withdifferent degrees of personal or corporate tone of voice whencommunicating with customers. This work also contributesby highlighting circumstances likely to explain the disparateresults of previous work about the effects of several factorsassociated with closeness and humanness in brand communica-tion (for example, Steinmann, Mau, and Schramm-Klein 2015;Van Noort and Willemsen 2012). It is possible that differentconsumption situations in previous work were major condi-tioners of their results. As suggested by our findings, the use ofhuman or corporate voice can interact with the consumer goalsand the level of situational involvement. It is even possible thatthe tone of voice interacts with other consumption attributes notexplored in this research that might have been a limitation ofprevious work's findings.

In addition, since the definition of human voice associateswith social presence theory, our findings can also be comparedto others following this theory to explain the effects of socialinteractions in e-commerce (e.g., Labrecque 2014). In the sameway, the concept of human voice is also related to brandhumanization in the sense that a brand using human voice isperceived as even more “human”. Therefore, our findings arealso relevant to studies on the anthropomorphism of brands(e.g., Puzakova, Kwak, and Rocereto 2013), specificallyresponding to the effects of humanized communication.

Managerial Implications

The results of this work also have implications for managers.Even though each study used the same product or service for themanipulations of consumer goals and situational involvement(to minimize confounded effects), the product or service offeredby a given brand may more usually be associated with a specificconsumption context. In other words, the products or services


offered by the brand may be purchased more often to satisfy ahedonic or utilitarian goal and be more associated with a low orhigh level of situational involvement. Hence, some brandsshould benefit more by using a more human voice, while othersshould use a corporate voice. Because of this, we can presentgeneral guidelines for companies about the tone of voice moreappropriate for their context (see Fig. 8).

Conditions Under Which Brands Should Use a Human Voiceon Social Media

It is a good idea for a brand to use a human voice in itsinteractions with customers on social media if its products orservices are primarily hedonic and associated with contexts oflow situational involvement and risk. This should be the casefor bookstores, travel agencies, sports clubs, etc. In addition, inthis condition, the majority of posts about the brand should notbe negative.

Conditions Under Which Use of a Human Voice on SocialMedia Makes Little Difference for Brands

If the brand's products or services are primarily utilitarian or ifoverall posts about the brand are negative, the benefits of using ahuman voice are limited or negligible. However, the brand maystill choose to use either a human or corporate voice in thesesituations, provided that the products or services are associatedwith contexts of low situational involvement and risk. The choiceof tone of voice should therefore follow what is most relevant interms of brand positioning. Examples include technical schools,cleaning services, and many business-to-business services.

Conditions Under Which Brands Should Not Use a HumanVoice on Social Media

It is not advisable for brands to use a human voice whentheir products or services are usually associated with contextsof high situational involvement and risk. In this case, it is betterto adopt a corporate voice and maintain a certain distance whendealing with customers. Financial consulting, medical services,and real estate brokers are examples in this category.

These recommendations are especially relevant for servicesmarketing, because of the importance of attendance andperipheral attributes in the customer experience (Bitner,

The brand’s products and services are hedonic, assocontexts of low situational involvement and risk, and

about the brand are not negative.

The brand’s products and services are associated withigh situational involvement and risk.

Overall posts about the brand on social media are

The brand’s products and services are utilitarian andwith contexts of low situational involvement an

Fig. 8. Framework for decisions about using a h

Booms, and Tetreault 1990). They should also be relevant forexperiential products, since customers are more inclined tosearch for the opinions of other people in this situation (Sénécaland Nantel 2004). Moreover, manipulations in human voice arepossible on most social media, even though at a smaller levelfor some (e.g., the space for brand communication is morelimited on TripAdvisor brand pages than it is on Facebook).Therefore, the findings of this paper, although extendable tosocial media in general, should be more useful for those inwhich the brand has more control over its contents.

Limitations and Further Research

In this paper, we focused on two attributes of the con-sumption context that moderate the influence of human voiceon purchase intentions: the consumer's goal (hedonic orutilitarian) and level of situational involvement (low or high).However, other factors might also interact with the use of ahuman voice and make its influence positive, negative, or nulltowards the brand. For example, a tone of voice may elicitdifferent expectations from the consumer depending on thebrand's positioning, typicality, perceived authenticity, and soon. Future studies may explore other mechanisms throughwhich tone of voice on social media influences consumerattitudes and behaviors. Similarly, this work focused on onesocial medium only – Facebook – since it is presently the mostused social medium for consumers and the most relevant formarketing (eMarketer 2016). Future studies could extend thepresent findings by testing the effects of human voice ondifferent social media in which the differences in the brandvoice might be less evident (such as Instagram) or in personalcommunications (email).

The studies in this research were restricted to the inves-tigation of the consumer's first contact with a new brand. Whilethis constituted a contribution to the literature, it also raisesthe question of the effects of tone of voice after repeatedinteractions, such as in an extended relationship. Should theeffects of using a human voice increase, tail off, or decreasewith time? Another question concerns the effects of adopting ahuman voice on consumers already familiar with the brand.Could a sudden change in tone of voice also change existing

It is advisable to use ahuman voice on social

media.

ciated withoverall posts

It is advisable to use acorporate voice on

social media.h contexts of

The choice of tone ofvoice makes little or

no difference.negative.

associatedd risk.

uman or a corporate voice on social media.


perceptions about the brand? Such an investigation should beparticularly interesting for brands dealing with service failures.Further research, therefore, could also investigate the choice ofa tone of voice to deal most effectively with complaints andunsatisfied customers.

Moreover, there are limitations to the methodologicalchoices of this research. Although the performed studies reliedon actual consumers, they used fictitious brands. Thus, studiesbased on consumer interactions with existing brands would bevaluable to improve the robustness of the results andinvestigate additional boundary conditions. Further research

Table A.1Dependent variables, independent variables and covariates a in Study 1a.

Variable Measurement items

Purchase intentions How likely would you stay at this hotel for a [(1 to 7: very unlikely–very likely)

Hedonic value (based upon Babin,Darden, and Griffin 1994 andPöyry, Parvinen, andMalmivaara 2013)

Regarding your experience with the hotel page… I enjoyed passing the time on the hotel page… Visiting the hotel page was a pleasant exper… Compared to the other things I could have d… I enjoyed visiting the hotel page for its own(1 to 7: totally disagree–totally agree)

Enduring involvement(Zaichkowsky 1985)

In your personal perceptions, hotels are… (1 to 7: important–unimportant)… (1 to 7: mean a lot to me–mean nothing to… (1 to 7: matter to me–do not matter to me)… (1 to 7: significant–insignificant)… (1 to 7: of concern to me–of no concern to… (1 to 7: interesting–boring)

Product expertise (based uponCai and Xu 2011)

Compared to other people, how familiar do yoDo you usually know precisely what attributesDo you think you can you make a satisfactorywithout another person's help? (1 to 7: never–

Facebook usage intensity (Ellison,Steinfield, and Lampe 2007)

Regarding your Facebook usage… Facebook is part of my everyday activity… I am proud to tell people I'm on Facebook… Facebook has become part of my daily rout… I feel out of touch when I haven't logged on… I feel I am part of the Facebook community… I would be sorry if Facebook shut down(1 to 7: totally disagree–totally agree)

a Gender, age, and education were also measured as covariates.

Appendix A

Table A.2Factor reliability and discriminant validity a for variables included in Study 1a.

Variable CR AVE Purchase intentions Hedonic valu

Purchase intentions .59 .59 .77Hedonic value .88 .65 .53 .81Enduring involvement .94 .73 −.35 −.51Product expertise .78 .55 .26 .36Facebook usage intensity .90 .61 .32 .40

a The numbers on the diagonal represent the square root AVE of each construct.factors.

investigating other industries, such as B2B, can also helpextend the external validity of the research. Finally, there is alimitation related to the characteristics of the participants of thisresearch (all from the United States).What would be the results ofthe experiments when applied to participants of other cultures? Itis a well-known fact that some cultures are more used touncertainty and closeness/distance than others (Hofstede 2001).Even within the U.S., some market segments may be more opento close and personal communications than others. Cross-culturalstudies should be of value to generalize the applicability of thesefindings.

Factorloadings

Alpha

vacation/business] trip to New York? .77 –

on Facebook .89.85

ience .78one, being on the hotel page was truly enjoyable .79sake, not just for the useful information I found .81

.95.88

me) .84.89.89

me) .86.77

u think you are with hotels? (1 to 7: very little–very much) .72 .72of a hotel decide its benefits? (1 to 7: never–all the time) .69choice of a hotel based on only your own knowledge,always)

.80

.89.80.67

ine .86to Facebook for a while .76

.78

.80

e Enduring involvement Product expertise Facebook usage intensity

.85−.47 .74−.36 .31 .78

Numbers below the diagonal depict the correlation of each factor with all other

Table A.3Manipulation checks in Study 1a.

Variable Measurement items Alpha

Humanness in tone of voice(based upon Kumar and Benbasat 2002)

Regarding your perceptions about the hotel page on Facebook .91… The [brand] created a sense of closeness with its audience… I felt close to [brand]… The [brand] created a sense of distance… I felt that the [brand] was aloof in its interactions with its audience… I found the [brand] to be very detached from its audience… The [brand] was very impersonal in its dealings with its audience… I found the [brand] to be very detached in its interactions with its audience(1 to 7: totally disagree–totally agree)

Consumer goal (Kronrod and Danziger 2013) In your opinion, the type of benefits offered by a hotel in a [business/vacation] trip is:(1 to 7: completely utilitarian–completely hedonic)

–

Realism How realistic do you think was the proposed situation?(1 to 7: very unrealistic–very realistic)

–

Role-playing How easy was it to imagine yourself in the proposed situation?(1 to 7: very difficult–very easy)

–

Fig. A.1. Extracts from brand pages and post examples in Study 1a.(These pictures are only extracts. The participants looked at brand pages with 29 posts each.)


Appendix B

Table B.1Dependent variables, independent variables and covariates a in Study 2a.

Variable Measurement items Factor loadings Alpha

Purchase intentions How likely would you go to this restaurant for [a quick dinner alone/a birthday dinner withyour family and friends]?(1 to 7: very unlikely–very likely)

.76 –

Perceived risk (based uponCampbell and Goodstein 2001)

The decision to go to Francesco's Pizza & Pasta for [a quick dinner alone/a birthday dinnerwith your family and friends] would make you feel

.92

… (1 to 7: not at all concerned–highly concerned) .93… (1 to 7: not at all worried–very worried)The decision to go to Francesco's Pizza & Pasta for [a quick dinner alone/a birthday dinnerwith your family and friends] would be

.94

… (1 to 7: not at all risky–extremely risky) .87Facebook usage intensity(Ellison, Steinfield, and Lampe 2007)

Regarding your Facebook usage .89… Facebook is part of my everyday activity .79… I am proud to tell people I'm on Facebook .79… Facebook has become part of my daily routine .80… I feel out of touch when I haven't logged onto Facebook for a while .71… I feel I am part of the Facebook community .85… I would be sorry if Facebook shut down(1 to 7: totally disagree–totally agree)

.79

Experience with restaurants Excluding vacation time, in average how many times per month you have dinner in a restaurant? .95 –

a Gender, age, and education were also measured as covariates.

Table B.2Factor reliability and discriminant validity a for variables included in Study 2a.

Variable CR AVE Purchase intentions Perceived risk Facebook usage intensity Experience with restaurants

Purchase intentions .58 .58 .76Perceived risk .94 .83 −.63 .91Facebook usage intensity .91 .62 .06 .04 .79Experience with restaurants .90 .90 .09 .03 .14 .95a The numbers on the diagonal represent the square root AVE of each construct. Numbers below the diagonal depict the correlation of each factor with all other

factors.

Table B.3Manipulation checks in Study 2a.

Variable Measurement items Alpha

Humanness in tone of voice (based upon Kumar and Benbasat2002)

Regarding your perceptions about the hotel page on Facebook .91… The [brand] created a sense of closeness with its audience… I felt close to [brand]… The [brand] created a sense of distance… I felt that the [brand] was aloof in its interactions with its audience… I found the [brand] to be very detached from its audience… The [brand] was very impersonal in its dealings with its audience… I found the [brand] to be very detached in its interactions with its audience(1 to 7: totally disagree–totally agree)

Situational involvement (based upon Mittal 1989 and Okazaki,Navarro-Bailón, and Molina-Castillo 2012)

In selecting from many restaurants available in the city for the situation described, wouldyou say that you would not care as to which one you choose?(1 to 7: I would not care at all–I would care a great deal)

.86

For the situation described, how important would it be for you to make the right choice for arestaurant?(1 to 7: not at all important–extremely important)In making your selection of a restaurant for the situation described, how concerned wouldyou be about the outcome of your choice?(1 to 7: not at all concerned–very much concerned)

Realism How realistic do you think was the proposed situation?(1 to 7: very unrealistic–very realistic)

–

Role-playing How easy was it to imagine yourself in the proposed situation?(1 to 7: very difficult–very easy)

–


Fig. B.1. Extracts from brand pages and post examples in Study 2a.(These pictures are only extracts. The participants looked at brand pages with 16 posts each.)


Appendix C

Fig. C.1. Extracts from brand pages in Study 2b.(These pictures are only extracts. The participants looked at brand pages with 11 posts each.)


References

Babin, Barry J., William R. Darden, and Mitch Griffin (1994), “Work and/or Fun:Measuring Hedonic and Utilitarian Shopping Value,” Journal of ConsumerResearch, 20, 4, 644–57.

Batra, Rajeev and Olli T. Ahtola (1990), “Measuring the Hedonic and UtilitarianSources of Consumer Attitudes,” Marketing Letters, 2, 2, 159–70.

Berry, Leonard L., Lewis P. Carbone, and Stephan H. Haeckel (2002), “Managingthe Total Customer Experience,”MIT SloanManagement Review, 43, 3, 84–9.

Beukeboom, Camiel J., Peter Kerkhof, and Metten de Vries (2015), “Does aVirtual Like Cause Actual Liking? How Following a Brand's FacebookUpdates Enhances Brand Evaluations and Purchase Intention,” Journal ofInteractive Marketing, 32, 26–36.

Bitner, Mary J., Bernard H. Booms, and Mary S. Tetreault (1990), “The ServiceEncounter: Diagnosing Favorable and Unfavorable Incidents,” Journal ofMarketing, 54, 1, 71–85.

Brown, Gavin and Gillian Maxwell (2002), “Customer Service in UK CallCentres: Organisational Perspectives and Employee Perceptions,” Journalof Retailing and Consumer Services, 9, 6, 309–16.

Cai, Shun and Yunjie Xu (2011), “Designing not Just for Pleasure: Effects ofWeb Site Aesthetics on Consumer Shopping Value,” International Journalof Electronic Commerce, 15, 4, 159–88.

Campbell, Margaret C. and Ronald C. Goodstein (2001), “The ModeratingEffect of Perceived Risk on Consumers' Evaluations of ProductIncongruity: Preference for the Norm,” Journal of Consumer Research,28, 3, 439–49.

Carlson, Kurt A., Margaret G. Meloy, and J. Edward Russo (2006), “Leader-driven Primacy: Using Attribute Order to Affect Consumer Choice,”Journal of Consumer Research, 32, 4, 513–8.

Childers, Terry L., Christopher L. Carr, Joann Peck, and Stephen Carson(2001), “Hedonic and Utilitarian Motivations for Online Retail ShoppingBehavior,” Journal of Retailing, 77, 4, 511–35.

http://refhub.elsevier.com/S1094-9968(17)30057-9/rf0010






























Choi, Jaewon, Hong Joo Lee, and Yong Cheol Kim (2011), “The Influence ofSocial Presence on Customer Intention to Reuse Online RecommenderSystems: The Roles of Personalization and Product Type,” InternationalJournal of Electronic Commerce, 16, 1, 129–54.

Dawson, Scott, Peter H. Bloch, and Nancy M. Ridgway (1990), “ShoppingMotives, Emotional States, and Retail Outcomes,” Journal of Retailing, 66,4, 408–27.

Dowling, Grahame R. and Richard Staelin (1994), “A Model of Perceived Riskand Intended Risk-handling Activity,” Journal of Consumer Research, 21,1, 119–34.

Ellison, Nicole B., Charles Steinfield, and Cliff Lampe (2007), “The Benefits ofFacebook ‘Friends:’ Social Capital and College Students' Use of OnlineSocial Network Sites,” Journal of Computer-Mediated Communication, 12,4, 1143–68.

eMarketer (2016), “Social Media Platforms Used by US Enterprise*Social Media Marketers, April 2016 (% of Respondents),” (July),(accessed July 13, 2016), [available at http://totalaccess.emarketer.com/Chart.aspx?R=194184].

Eroglu, Sevgin A., Karen A. Machleit, and Lenita M. Davis (2003), “EmpiricalTesting of a Model of Online Store Atmospherics and Shopper Responses,”Psychology and Marketing, 20, 2, 139–50.

Fournier, Susan (1998), “Consumers and their Brands: Developing RelationshipTheory in Consumer Research,” Journal of Consumer Research, 24, 4,343–73.

Gretry, Anaïs, Csilla Horváth, Nina Belei, and Allard C.R. van Riel (2017),“‘Don't Pretend to Be my Friend!’ When an Informal Brand Communi-cation Style Backfires on Social Media,” Journal of Business Research, 74,77–89.

Hassanein, Khaled and Milena Head (2006), “The Impact of Infusing SocialPresence in the Web Interface: An Investigation Across Product Types,”International Journal of Electronic Commerce, 10, 2, 31–55.

Hayes, Andrew F. (2013), Introduction to Mediation, Moderation, andConditional Process Analysis: A Regression-based Approach. New York:Guilford Press.

Hirschman, Elizabeth C. andMorris B. Holbrook (1982), “Hedonic Consumption:Emerging Concepts, Methods and Propositions,” Journal of Marketing, 46, 2,92–101.

Hofstede, Geert (2001), Culture's Consequences: Comparing Values, Behaviors,Institutions, and Organizations Across Nations. London: Sage.

Houston, Michael J. and Michael L. Rothschild (1978), “Conceptual andMethodological Perspectives on Involvement,” in Research Frontiers inMarketing: Dialogues and Directions. Subhash C. Jain, editor. Chicago:American Marketing Association. 184–7.

Huang, Ming-Hui (2006), “Flow, Enduring, and Situational Involvement in theWeb Environment: A Tripartite Second-order Examination,” Psychologyand Marketing, 23, 5, 383–411.

Keeling, Kathleen, Peter McGoldrick, and Susan Beatty (2010), “Avatars asSalespeople: Communication Style, Trust, and Intentions,” Journal of BusinessResearch, 63, 793–800.

Kelleher, Tom (2009), “Conversational Voice, Communicated Commitment,and Public Relations Outcomes in Interactive Online Communication,”Journal of Communication, 59, 1, 172–88.

Kronrod, Ann and Shai Danziger (2013), “‘Wii Will Rock you!’ TheUse and Effect of Figurative Language in Consumer Reviews of Hedonicand Utilitarian Consumption,” Journal of Consumer Research, 40, 4,726–39.

Kumar, Nanda and Izak Benbasat (2002), “Para-social Presence andCommunication Capabilities of a Web Site,” e-Service Journal, 1, 3, 5–24.

Labrecque, Lauren I. (2014), “Fostering Consumer–Brand Relationships inSocial Media Environments: The Role of Parasocial Interaction,” Journal ofInteractive Marketing, 28, 2, 134–48.

Li, Jui-Min, Jen-Shou Yang, and Hsin-Hsi Wu (2009), “Analysis of CompetencyDifferences Among Frontline Employees from Various Service Typologies:Integrating the Perspectives of the Organisation and Customers,” The ServiceIndustries Journal, 29, 12, 1763–78.

Liang, Ting-Peng and Efraim Turban (2011), “Introduction to the SpecialIssue Social Commerce: A Research Framework for Social Commerce,”International Journal of Electronic Commerce, 16, 2, 5–14.

Lund, Kevin and Eileen Sutton (2014), “Why Your Brand Should Speak Human,”Content Marketing Institute (accessed June 30, 2014), [available at http://http://contentmarketinginstitute.com/2014/11/why-brand-speak-human].

McKenna, Katelyn Y.A., Amie S. Green, and Marci E.J. Gleason (2002),“Relationship Formation on the Internet: What's the Big Attraction?”Journal of Social Issues, 58, 1, 9–31.

Merle, Aurélie, Sylvain Sénécal, and Anik St-Onge (2012), “Whetherand How Virtual Try-on Influences Consumer Responses to an ApparelWeb Site,” International Journal of Electronic Commerce, 16, 3,41–64.

Meyer, Kate (2016), “The Four Dimensions of Tone of Voice,” Nielsen NormanGroup (accessed June 10, 2017), [available at https://www.nngroup.com/articles/tone-of-voice-dimensions/].

Mittal, Banwari (1989), “Measuring Purchase-decision Involvement,” Psychologyand Marketing, 6, 2, 147–62.

Okazaki, Shintaro, María Ángeles Navarro-Bailón, and Francisco-Jose Molina-Castillo (2012), “Privacy Concerns in Quick Response Code MobilePromotion: The Role of Social Anxiety and Situational Involvement,”International Journal of Electronic Commerce, 16, 4, 91–120.

Park, Hyojung and Hyunmin Lee (2013), “Show Us You Are Real: The Effectof Human-Versus-Organizational Presence on Online Relationship BuildingThrough Social Networking Sites,” Cyberpsychology, Behavior and SocialNetworking, 16, 4, 265–71.

——— and Glen T. Cameron (2014), “Keeping it Real: Exploring the Roles ofConversational Human Voice and Source Credibility in Crisis Communi-cation via Blogs,” Journalism and Mass Communication Quarterly, 91, 3,487–507.

Payne, John W., James R. Bettman, and Eric J. Johnson (1993), The AdaptiveDecision Maker. Cambridge, UK: Cambridge University Press.

Peer, Eyal, Joachim Vosgerau, and Alessandro Acquisti (2014), “Reputation asa Sufficient Condition for Data Quality on Amazon Mechanical Turk,”Behavior Research Methods, 46, 4, 1023–31.

Pöyry, Essi, Petri Parvinen, and Tuuli Malmivaara (2013), “Can We Get fromLiking to Buying? Behavioral Differences in Hedonic and UtilitarianFacebook Usage,” Electronic Commerce Research and Applications, 12, 4,224–35.

Puzakova, Marina, Hyokjin Kwak, and Joseph F. Rocereto (2013), “WhenHumanizing Brands Goes Wrong: The Detrimental Effect of BrandAnthropomorphization amid Product Wrongdoings,” Journal of Marketing,77, 3, 81–100.

Roser, Connie (1990), “Involvement, Attention, and Perceptions of MessageRelevance in the Response to Persuasive Appeals,”Communication Research,17, 5, 571–600.

Schamari, Julia and Tobias Schaefers (2015), “Leaving the Home Turf: HowBrands Can Use Webcare on Consumer-generated Platforms to IncreasePositive Consumer Engagement,” Journal of Interactive Marketing, 30,20–33.

Sela, Aner, S. Christian Wheeler, and Gülen Sarial-Abi (2012), “We Are not theSame as you and I: Causal Effects of Minor Language Variations onConsumers' Attitudes Toward Brands,” Journal of Consumer Research, 39,3, 644–61.

Sen, Shahana and Dawn Lerman (2007), “Why Are You Telling Me This? AnExamination into Negative Consumer Reviews on the Web,” Journal ofInteractive Marketing, 21, 4, 76–94.

Sénécal, Sylvain, Jamel-Edine Gharbi, and Jacques Nantel (2002), “TheInfluence of Flow on Hedonic and Utilitarian Shopping Values,” Advancesin Consumer Research, 29, 1, 483–4.

——— and Jacques Nantel (2004), “The Influence of Online ProductRecommendations on Consumers' Online Choices,” Journal of Retailing,80, 2, 159–69.

Sheth, Jagdish N., Bruce I. Newman, and Barbara L. Gross (1991),ConsumptionValues and Market Choices: Theory and Applications. Cincinnati, OH:Southwestern Publishing.

Short, John, EderynWilliams, and Bruce Christie (1976), The Social Psychologyof Telecommunications. London, UK: John Wiley & Sons.

Snyder, Benjamin (2015), “Facebook Added 10 Million Small Business Pages ina Year,” Fortune (accessed April 30, 2015), [available at http://fortune.com/2015/04/30/facebook-small-business/].















http://totalaccess.emarketer.com/Chart.aspx?R=194184

http://totalaccess.emarketer.com/Chart.aspx?R=194184


















































http://http://contentmarketinginstitute.com/2014/11/why-brand-speak-human

http://http://contentmarketinginstitute.com/2014/11/why-brand-speak-human







https://www.nngroup.com/articles/tone-of-voice-dimensions/

https://www.nngroup.com/articles/tone-of-voice-dimensions/




















































http://fortune.com/2015/04/30/facebook-small-business/

http://fortune.com/2015/04/30/facebook-small-business/


Statista (2017a), “Number of Social Media Users Worldwide from 2010 to2020,” (accessed June 10, 2017), [available at https://www.statista.com/statistics/278414/number-of-worldwide-social-network-users].

——— (2017b), “Social Media Advertising: Trends, Insights & Top Players,”(accessed June 10, 2017), [available at https://www.statista.com/outlook/220/100/social-media-advertising/worldwide#market-revenue].

Steinmann, Sascha, Gunnar Mau, and Hanna Schramm-Klein (2015), “BrandCommunication Success in Online Consumption Communities: AnExperimental Analysis of the Effects of Communication Style and BrandPictorial Representation,” Psychology and Marketing, 32, 3, 356–71.

Van der Heijden, Hans (2004), “User Acceptance of Hedonic InformationSystems,” MIS Quarterly, 28, 4, 695–704.

Van Noort, Guda and Lotte M. Willemsen (2012), “Online Damage Control:The Effects of Proactive Versus Reactive Webcare Interventions inConsumer-generated and Brand-generated Platforms,” Journal of InteractiveMarketing, 26, 3, 131–40.

Verhagen, Tibert, Jaap van Nes, Frans Feldberg, and Willemijn van Dolen(2014), “Virtual Customer Service Agents: Using Social Presence andPersonalization to Shape Online Service Encounters,” Journal of Computer-Mediated Communication, 19, 3, 529–45.

Wang, Liz C., Julie Baker, Judy A. Wagner, and Kirk Wakefield (2007), “Can aRetail Web Site Be Social?” Journal of Marketing, 71, 3, 143–57.

Zaichkowsky, Judith L. (1985), “Measuring the Involvement Construct,”Journal of Consumer Research, 12, 3, 341–52.

Zhang, Jing and En Mao (2008), “Understanding the Acceptance of MobileSMS Advertising Among Young Chinese Consumers,” Psychology andMarketing, 25, 8, 787–805.

Renato Hübner Barcelos is Postdoctoral Fellow at HEC Montreal. Hisinterests in research are related to digital marketing, consumer behavior andmarketing on social media. He has taught marketing and consumer behaviorcourses in Canada and Brazil. He holds a M.Sc. and a Ph.D. in marketing fromFederal University of Rio Grande do Sul (Brazil).

Danilo Correa Dantas is Associate Professor at HEC Montreal. His researchinterests are related to database marketing, music marketing and electroniccommerce. His work has been published in marketing and managementscientific journals such as Management Decision and Canadian Journal ofAdministrative Sciences. He holds a Ph.D. in marketing from Université PierreMendès-France (Grenoble II).

Sylvain Sénécal is Professor of Marketing, RBC Financial Group E-commerceChair, and Tech3Lab Co-director at HECMontreal. His research interests are relatedto consumer marketing on the Internet and consumer neuroscience. His work hasbeen published inmarketing and ecommerce scientific journals such as the Journal ofthe Academy of Marketing Science and the International Journal of ElectronicCommerce. He holds a M.Sc. and a Ph.D. in marketing from HEC Montreal.

https://www.statista.com/statistics/278414/number-of-worldwide-social-network-users

https://www.statista.com/statistics/278414/number-of-worldwide-social-network-users

[email protected]

[email protected]





















Watch Your Tone: How a Brand's Tone of Voice on Social ...€¦ · HEC Montreal, 3000 Chemin de la Côte-Sainte-Catherine ... Online branding; Digital marketing Introduction The popularity

Documents